En cuanto a la historia

The aim of this thesis was to demonstrate the feasibility of producing a selective stimulator using VLSI techniques. Although stimulators have been realised using integrated circuits in the past they have tended to rely on digital electronics to produce the stimulation waveforms. Due to the quasi-trapezoidal stimulation waveforms required for selective stimulation, analogue circuit techniques were deemed more appropriate in this stimulator design. In this study, the modules for a selective stimulator unit were implemented using Mietec 2.4|im CMOS technology. The principle results of the work presented in this thesis will now be summarised and discussed.

Following the general introduction given in Chapter 1, Chapter 2 reviews the history of selective stimulation, discussing the types of waveform and nerve cuffs that have being used for selection by fibre size and selection by fibre position. Conclusions are drawn from the literature and each section ends with a table showing the different methods used for obtaining selective stimulation. This review of selective stimulation methods has not being reported in the literature previously.

Chapter 3 presents the development of the specification for a selective stimulator system. The summaries developed for selective stimulation by fibre size and fibre position were invaluable in the development of the specification for a selective stimulator system. The specification allows the new device to control three different types of nerve electrode cuffs (dipoles, tripoles and pentapoles). The tripoles (and dipoles) are to be used for selection by fibre size. The specification introduces a new cuff design, the pentapole that consists of four central ‘dot’ electrodes and two anode rings. It is hoped that this cuff design, used with the new stimulator device, will, in practice, allow chronic investigation of selective stimulation by fibre position. The pentapolar electrode cuff may also allow the investigation of selective stimulation by fibre position in conjunction with selective stimulation by fibre size, a technique that has not yet being investigated. However, as the blocking of the nerve fibres occurs at the anodes the design of the nerve cuff may have to be modified to contain a central stimulating cathode and multiple anodes. This possibility will require acute and chronic investigations.

The output currents of the stimulator to each cuff are set as ratios, either of cathode currents in the pentapole or anode currents in the tripole. The specification proposes that the stimulator produce charge-balanced stimulation currents of variable duration.

C hapter 6_______________________________________Summary______________________________________________177

amplitude, and reverse amplitude with and without an exponential tail. This greatly exceeds the number of parameters controlled by most existing stimulators.

In Chapter 4 the design, fabrication and testing results of the circuits required for a selective stimulator are discussed. The modules were first fabricated on three test ICs, altogether occupying more than 35mm^ of silicon.

A novel transconductance circuit is presented that allows ±5mA current pulses to be generated with loads approaching IkO impedance. The simulation results showed that the circuit behaves as a linear transconductor with a variable transconductance defined by a single bias current. The nominal transconductance for the stimulator was 3.5mA/V. The transconductor demonstrated, in simulation, a 1.5% maximal non-linearity (at 4mA) with a 10% mismatch in transistor parameters. This is approximately double the non- linearity demonstrated by the Park and Schununan transconductor [99,100], however the output voltage swing of the developed transconductor is over double that achievable with their circuit. The experimental results highlighted a problem with the common-mode feedback that caused the circuit to behave unstably. This problem was analysed in simulation and a modified feedback scheme was proposed. The modified feedback scheme was implemented in a later integrated circuit and allowed the transconductor to operate stably. The offset of the developed differential transconductor was shown in simulation to be close to zero. The effective differential offset of this device is defined by the differential offset presented at its inputs (ref to section).

A long time constant integrator (x=lms, fs=250kHz) has been developed that is very area-efficient compared to other types of integrator investigated; requiring only 30% of the area required by a classical switched capacitor integrator with the same time constant. This circuit was a development of a low gain integrator proposed by Sansen and van Peteghem [125,147]. The capacitance spread is reduced from 200:1 to 15:1, making the circuit realisation and capacitor matching easier. The DC feedback method presented allows the low frequency gain of the integrator to be simply defined. The analysis shown for this circuit developed simple equations for both the time constant and the gain of this integrator. The experimental results from this circuit highlighted a problem with charge injection sensitivity in some of the input nodes of the circuit. The solution to this problem is to use dummy switches to cancel the charge injection in the critical nodes. With the proposed modifications, the DC offset voltage of the integrator was reduced

C h apter 6_______________________________________Summ ary______________________________________________178

from an experimental (and later simulated) offset of 227mV to a simulated offset of around 7mV.

A pulse-generator circuit, capable of generating the waveform shapes required for selective stimulation of definable amplitude, duration, reverse amplitude and tail shape was developed. The circuits performance was shown to meet the specification developed for the circuit with the exception that the reverse amplitude ratio was slightly too low at the smallest ratios (2%). This should be corrected in future devices.

The other circuits for the stimulator unit all functioned within the expected tolerances when tested. Overall, the design presented was suitable for integrated circuit techniques, except for the external capacitor and resistor required by the pulse generation circuit. Two external components for the whole stimulation unit is acceptable when compared with the fact that every stimulating electrode requires a blocking capacitor of several jiF, each of which will be of a volume comparable to the IC.

The overall offset needs careful consideration in a complete stimulator system. Because negligible offset is due to the transconductors (see above), the effective offset at the output is the transconductance multiplied by differential offset voltage presented to the transconductors. This offset voltage is effectively {Vojfset (DAC) + Vojfset (Pulse- Generator). + Vq#j€f(differentiai) (Attenuators)). The experimental results presented in 4.10 showed that this offset voltage will be below 5|xA when the proposed improvements to reduce the offset of the attenuator amplifiers have been implemented. This equates to a maximal charge error^^ of 4% (0.2pC) of the maximal pulse charge. This charge has to be removed by the discharge array. Future designs may consider the addition of offset correction at the transconductor outputs to improve the overall accuracy of the charge balancing.

Chapter 5 firstly proposes applications for the stimulator; these applications include an improved stimulator for bladder and bowel and a foot-drop stimulator. The ability to generate unidirectionally-propagating action potentials should allow the device to be used with a larger group of patients. Secondly, Chapter 5 discusses the work required to turn the test devices, discussed in Chapter 4, into a complete selective stimulator system. This system has a minimal number of internal components: voltage regulators, handshaking transistors, an RF filter (to extract the transmitted data and power), a digital control IC, a stimulation unit, blocking capacitors plus a few other discrete components.

C hapter 6_______________________________________Summary______________________________________________179

This minimal number of components means that the stimulator can be physically small, allowing the device to be placed closer to the stimulation site.

In conclusion this thesis presents the development of a stimulation IC, which, when incorporated into a complete stimulator system, will allow selective stimulation techniques to be investigate in chronic human clinical trials.

The claims of this thesis are as follows.

1. A detailed literature review of selective stimulation using implanted nerve cuff electrodes was presented and summaries of both stimulation methods were presented. This detailed review has not appeared in the literature before. 2. A specification for a selective stimulator was developed from the information

generated by the literature review. The specification allows control of pulse amplitude, duration, reverse current amplitude and tail time-constant.

3. Circuits suitable for a selective stimulator were developed, simulated, and tested experimentally. Where necessary improvements to the realised designs have been proposed. The circuits developed for the stimulator are listed in 4- 9.

4. An 8-bit DAC based mostly on standard cells.

5. A pulse-generator circuit, capable of generating the waveform shapes required for selective stimulation of definable amplitude, duration, reverse amplitude and tail shape.

6. A four-channel, single-sided to differential attenuator.

7. A novel linear differential transconductor capable of delivering ± 5mA into a 1 kohm load with a low offset current and using only ± 5V power supplies. The offset of the transconductor is effective defined by differential offset voltages presented at the input of the transconductor.

8. Area efficient switching and discharge switching arrays.

9. A long time-constant integrator with increased DC gain has beenn developed from a design proposed by Sansen and van Petegham. The circuit has being simulated, analysed and further improvements suggested.

10. The overall performance of the circuits (with some modifications) will allow a complete stimulator system to be developed using these designs.

C h apter 6_______________________________________Summary______________________________________________180

11. Applications for the stimulator have being proposed.

12. The developments required to produce a complete stimulator system have being discussed.

Appendix I Testing o f the Integrated Circuits 181